loading model clobbers existing logging configuration

[rkechols]

Search before asking

• I have searched the YOLOv5 issues and found no similar bug report.

YOLOv5 Component

PyTorch Hub

Bug

Loading this repository using torch.hub.load results in existing logger configurations being partially clobbered. Specifically, upon loading this repository, the root logger has a new handler added to it which prints the log message to STDERR with no formatting. Additionally, the root logger level is set to either INFO or ERROR (depending on parameters and environment variables).

Environment

• YOLOv5 🚀 v6.2-237-g55e9516 Python-3.9.11 torch-1.12.1 CPU
• OS: macOS 13.0 (22A380)

Minimal Reproducible Example

import logging
import logging.config
import torch

logging.config.dictConfig({
    "version": 1,
    "formatters": {
        "formatter_default": {
            "format": "%(asctime)s - %(levelname)8s - %(name)s - %(message)s"
        },
    },
    "handlers": {
        "handler_stdout": {
            "class": "logging.StreamHandler",
            "formatter": "formatter_default",
            "level": "DEBUG",
            "stream": "ext://sys.stdout",
        },
    },
    "loggers": {
        "my.unrelated.logger": {
            "level": "DEBUG"
        },
        "yolov5": {
            "level": "CRITICAL"
        },
    },
    "root": {
        "level": "INFO",
        "handlers": [
            "handler_stdout",
        ],
    },
})
UNRELATED_LOGGER = logging.getLogger("my.unrelated.logger")


def log_test(message: str):
    UNRELATED_LOGGER.debug(message)
    UNRELATED_LOGGER.info(message)
    UNRELATED_LOGGER.warning(message)
    UNRELATED_LOGGER.error(message)
    UNRELATED_LOGGER.critical(message)
    for handler_ref in logging._handlerList:
        handler_ref().flush()


log_test("before torch.hub.load")

model = torch.hub.load(
    "ultralytics/yolov5",  # commit tested: 55e95168465b094733e3ef1ec36e0a18f200cd94
    "yolov5s",
    trust_repo=True,
    pretrained=False,  # just download and load the repo; no specific weights needed
    device="cpu",
    verbose=False,
)

log_test("after torch.hub.load")

Note that above produces the following to STDOUT and STDERR respectively:

STDOUT:

2022-11-11 10:46:00,781 -    DEBUG - my.unrelated.logger - before torch.hub.load
2022-11-11 10:46:00,782 -     INFO - my.unrelated.logger - before torch.hub.load
2022-11-11 10:46:00,782 -  WARNING - my.unrelated.logger - before torch.hub.load
2022-11-11 10:46:00,782 -    ERROR - my.unrelated.logger - before torch.hub.load
2022-11-11 10:46:00,782 - CRITICAL - my.unrelated.logger - before torch.hub.load
2022-11-11 10:46:02,888 -    DEBUG - my.unrelated.logger - after torch.hub.load
2022-11-11 10:46:02,888 -     INFO - my.unrelated.logger - after torch.hub.load
2022-11-11 10:46:02,888 -  WARNING - my.unrelated.logger - after torch.hub.load
2022-11-11 10:46:02,888 -    ERROR - my.unrelated.logger - after torch.hub.load
2022-11-11 10:46:02,888 - CRITICAL - my.unrelated.logger - after torch.hub.load

STDERR:

after torch.hub.load
after torch.hub.load
after torch.hub.load
after torch.hub.load

According to my configuration for my.unrelated.logger, STDERR should not receive any messages from my.unrelated.logger.

Additional

Separate from this actual bug is the fact that the repo is trying to set logging configurations in imported code. Logging configurations should only be set by the python program's direct entrypoint. Setting logging configurations in packages, libraries, imported modules, etc. removes logging control from top-level code.

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[github-actions]

👋 Hello @rkechols, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://ultralytics.com or email support@ultralytics.com.

Requirements

Python>=3.7.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training, validation, inference, export and benchmarks on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

@rkechols got it, thank you! I see your PR, I'll move the convo there.

[rkechols]

Related PR fix (#10133) has been merged.

[rkechols]

Related PR fix (#10192) has been merged.

[glenn-jocher]

@rkechols awesome, thanks for letting me know! If you need any further assistance, feel free to reach out.